|
PureBytes Links
Trading Reference Links
|
Bob,,
At the risk of repeaing myself, I am a french teacher in physics and
chemistry, graduate from the university and my teacher diplom is for real.
Me an my colleagues have learnt the same things at the university and
teached the same things and we do not invent anything.
It seems that you still confuse mathematical calculus with infinite
precision with experimental calculus that has it owns mathematical rules
when dealing with the precision display that we use in experimental science.
A numerical number like 2 is written 2 in mathematics,ans has an infinite
precision ( decimal are aditted and infinite suite of zeros)
In physics ( where this number comes from a measure) it is written
2.*******...***
where * is a decimal known for sure.
If it is not known, it is not written.
When you write 2.00 it means that the measure result if for sure 2.00, or if
you prefer 2.00xxxxxxxxxxxx
where xxxxxxxx is garbage.
We never write the garbage part.
Your example below assume that TS lose precision when differencing two close
numbers.
This is true for TS an for any software, even for a pencil and paper.
Differencing two close numbers implies a lost of precision, baeause the
difference sets to 0 the left digits and only leave the right digits.
What is sad is that you are maybe advocating that TS should replace the
left digits set to 0 ( that are no more significative digits) by other
digits far to the right, what you do not have the right to do, because these
digits can only be obtained by differencing the garbage part of the numbers.
In fact you draw this false conclusion when you compare the TS precision
10^-7 with the precision of the difference 10^-3, and regret that 10^-3 was
obtained by TS so that TS is guilty of that!
> So the TradeStation precision of one part in 10^7 in the ORIGINAL
> VALUES caused an error of 1.8 parts in 10^3 in the difference
> between the two values.
>
You say "TradeStation precision CAUSED"
NO, NO, NO, & No.
It is the original numbers precision (10^-6) and the close values of them
that caused the loss of precision in the difference!
What I regret, because all of the fist part of your demonstration was
perfectly correct
Either if I explain by simple example like this or by error calculus, it
seems that the few active posting members cannot understand this, including
you.
At least this was not the case for Patrick Gamble ( thanks).
Maybe there is something else on this list who understand the precision
calculus allowed display concept .
I cannot believe that it is only a french specificity...
Sincerely,
Pierre Orphelin
www.sirtrade.com
TradeStation Technologies representative in France
> -----Message d'origine-----
> De : Bob Fulks [mailto:bfulks@xxxxxxxxxxxx]
> Envoye : samedi 28 juillet 2001 22:08
> A : omega-list@xxxxxxxxxx
> Objet : RE: TradeStation Precision - Summary
>
>
> Interesting, but very fuzzy logic (pun untended, considering the
> source). I have no more time to waste arguing with you on this.
>
> YOU ARE WRONG!
>
> To explain it in extreme detail one more time, take an extremely
> simple case. Assume we have two values and we want to find the
> difference between the values.
>
> Assume that if we had perfect arithmetic, the two values would be:
>
> 10323.45026584125452... and 10321.20065325413...
>
> TradeStation will represent each number with precision of one part in
> 10^7 so TradeStation will represent these values as:
>
> 10323.450xxxx and 10321.200xxxx
>
> where the xxxx are numbers with no accuracy.
>
> So when we see this and since we do not know the true numbers, all we
> know is that the true value might be anywhere within a window of
> 0.001 on either side of the above values:
>
> 10323.449xxxx and 10321.199xxxx Minimum
> 10323.451xxxx and 10321.201xxxx Maximum
>
> So the difference between the two numbers can have values ranging
> from:
>
> 10323.449xxxx - 10321.201xxxx = 2.248xxxx Minimum
> 10323.451xxxx - 10321.199xxxx = 2.252xxxx Maximum
>
> So the maximum uncertainty in the difference is:
>
> Uncertainty: 0.004xxxx
>
> The uncertainty as a ratio to the value = 0.004 / 2.25 = 0.0018
>
> This is 1.8 parts in 10^3
>
> So the TradeStation precision of one part in 10^7 in the ORIGINAL
> VALUES caused an error of 1.8 parts in 10^3 in the difference
> between the two values.
>
> Certainly, TradeStation can represent the difference to seven
> significant places of precision:
>
> 2.248xxxx Minimum
> 2.252xxxx Maximum
>
> but all of the places shown by the "xxxx" have no accuracy because of
> the ERROR IN REPRESENTING THE ORIGINAL NUMBERS.
>
> You say:
>
> "...and that in any case, the TS precision do not damage the
> precision of the calculus."
>
> I repeat: YOU ARE WRONG!
>
> You clearly do not understand this and I have exhausted my methods of
> explaining it to you. If ANY of the thousand or so people on this
> list think Pierre is correct, please tell us...
>
> Bob Fulks
>
|